import pandas as pd
import warnings
warnings.filterwarnings('ignore')
import plotly.express as px

df = pd.read_excel("../ME22_Concentration_Particles_Raw data.xlsx")

df

	Sample	Particle Id	Group	Score	PM80-1 (µg/m3)	Particle # conc. (#particle/m3)	CmNormPM10-PM2.5 (µg/m3)	Mass deposition (µg/cm2/s)	Number deposition (#particles/cm2/s)	MassMicrogram (µg)	...	Th	Pa	U	Np	Pu	Am	Cm	Bk	Cf	Es
0	BO22_Car_p1_sec. analysis	16	Biogenic_Organic	0.961	0.048785	298.552352	0.042244	6.920258e-15	0.000042	0.000163	...	0	0	0	0	0	0	0	0	0	0
1	BO22_Car_p1_sec. analysis	22	Biogenic_Organic	0.972	0.042139	400.144760	0.038623	4.459923e-15	0.000042	0.000105	...	0	0	0	0	0	0	0	0	0	0
2	BO22_Car_p1_sec. analysis	130	Biogenic_Organic	0.963	0.030344	771.693922	0.028766	1.665273e-15	0.000042	0.000039	...	0	0	0	0	0	0	0	0	0	0
3	BO22_Car_p1_sec. analysis	132	Biogenic_Organic	0.963	0.023648	1270.575606	0.019901	7.882287e-16	0.000042	0.000019	...	0	0	0	0	0	0	0	0	0	0
4	BO22_Car_p1_sec. analysis	150	Biogenic_Organic	0.780	0.021368	1556.176408	0.015646	5.815202e-16	0.000042	0.000014	...	0	0	0	0	0	0	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
7415	ME22_TLY_P2	1446	Tire wear	0.973	0.035577	136.383237	0.023998	4.658701e-15	0.000032	0.000261	...	0	0	0	0	0	0	0	0	0	0
7416	ME22_TLY_P2	1500	Tire wear	0.989	0.024097	297.297857	0.022019	1.447500e-15	0.000032	0.000081	...	0	0	0	0	0	0	0	0	0	0
7417	ME22_TLY_P2	1614	Tire wear	0.384	0.029359	200.269512	0.024489	2.618087e-15	0.000032	0.000147	...	0	0	0	0	0	0	0	0	0	0
7418	ME22_TLY_P2	1662	Tire wear	0.964	0.017879	540.049822	0.016994	5.912288e-16	0.000032	0.000033	...	0	0	0	0	0	0	0	0	0	0
7419	ME22_TLY_P2	1701	Tire wear	0.801	0.016186	658.881086	0.015200	4.387284e-16	0.000032	0.000025	...	0	0	0	0	0	0	0	0	0	0

7420 rows × 144 columns

df.Group.value_counts().plot(kind='bar')

<AxesSubplot:>

df

	Sample	Particle Id	Group	Score	PM80-1 (µg/m3)	Particle # conc. (#particle/m3)	CmNormPM10-PM2.5 (µg/m3)	Mass deposition (µg/cm2/s)	Number deposition (#particles/cm2/s)	MassMicrogram (µg)	...	Th	Pa	U	Np	Pu	Am	Cm	Bk	Cf	Es
0	BO22_Car_p1_sec. analysis	16	Biogenic_Organic	0.961	0.048785	298.552352	0.042244	6.920258e-15	0.000042	0.000163	...	0	0	0	0	0	0	0	0	0	0
1	BO22_Car_p1_sec. analysis	22	Biogenic_Organic	0.972	0.042139	400.144760	0.038623	4.459923e-15	0.000042	0.000105	...	0	0	0	0	0	0	0	0	0	0
2	BO22_Car_p1_sec. analysis	130	Biogenic_Organic	0.963	0.030344	771.693922	0.028766	1.665273e-15	0.000042	0.000039	...	0	0	0	0	0	0	0	0	0	0
3	BO22_Car_p1_sec. analysis	132	Biogenic_Organic	0.963	0.023648	1270.575606	0.019901	7.882287e-16	0.000042	0.000019	...	0	0	0	0	0	0	0	0	0	0
4	BO22_Car_p1_sec. analysis	150	Biogenic_Organic	0.780	0.021368	1556.176408	0.015646	5.815202e-16	0.000042	0.000014	...	0	0	0	0	0	0	0	0	0	0
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
7415	ME22_TLY_P2	1446	Tire wear	0.973	0.035577	136.383237	0.023998	4.658701e-15	0.000032	0.000261	...	0	0	0	0	0	0	0	0	0	0
7416	ME22_TLY_P2	1500	Tire wear	0.989	0.024097	297.297857	0.022019	1.447500e-15	0.000032	0.000081	...	0	0	0	0	0	0	0	0	0	0
7417	ME22_TLY_P2	1614	Tire wear	0.384	0.029359	200.269512	0.024489	2.618087e-15	0.000032	0.000147	...	0	0	0	0	0	0	0	0	0	0
7418	ME22_TLY_P2	1662	Tire wear	0.964	0.017879	540.049822	0.016994	5.912288e-16	0.000032	0.000033	...	0	0	0	0	0	0	0	0	0	0
7419	ME22_TLY_P2	1701	Tire wear	0.801	0.016186	658.881086	0.015200	4.387284e-16	0.000032	0.000025	...	0	0	0	0	0	0	0	0	0	0

7420 rows × 144 columns

Check a histogram for the score value

Chemic

Chemical distributions of morphochemical groups

pd.set_option('display.max_columns', None)
mineral_df = df[df.Group == 'Mineral']

chemic_features = ["Si", "C", "O", "P", "Cl", "Na", "Ca", "Al", "Fe"]

columns_list = list(df.columns)
H_idx = columns_list.index('H')
mineral_chemic = mineral_df.iloc[:, H_idx:columns_list.index('Bi')]
mineral_chemic.loc[:, (df != 0).any(axis=0)].boxplot(figsize=(18, 6))

<AxesSubplot:>

metallic_df = df[df.Group == 'Metallic']
columns_list = list(df.columns)
H_idx = columns_list.index('H')
metallic_chemic = metallic_df.iloc[:, H_idx:columns_list.index('Bi')]
metallic_chemic.loc[:, (df != 0).any(axis=0)].boxplot(figsize=(18, 6))

<AxesSubplot:>

tire_df = df[df.Group == 'Tire wear']
columns_list = list(df.columns)
H_idx = columns_list.index('H')
tire_chemic = tire_df.iloc[:, H_idx:columns_list.index('Bi')]
tire_chemic.loc[:, (df != 0).any(axis=0)].boxplot(figsize=(18, 6))

<AxesSubplot:>

organic_df = df[df.Group == 'Biogenic_Organic']
columns_list = list(df.columns)
H_idx = columns_list.index('H')
organic_chemic = organic_df.iloc[:, H_idx:columns_list.index('Bi')]
organic_chemic.loc[:, (df != 0).any(axis=0)].boxplot(figsize=(18, 6))

<AxesSubplot:>

Composicion quimica de grupos

H_idx = columns_list.index('H')
#grp = df.iloc[:, H_idx:columns_list.index('Bi')]
#grp['Group'] = df['Group'].copy()
grp = df[chemic_features + ['Group']]
grp = grp.groupby('Group', group_keys=False).apply(lambda x: x.sample(50, replace=True)).copy()
vis = grp.copy()
vis.Group = vis.Group.astype('category').cat.codes
fig = px.parallel_coordinates(vis.reset_index().drop('index', axis=1), color = "Group",
                        range_color=[0, 5],color_continuous_scale=[(0.00, "grey"),   (0.16, "grey"),
                                                (0.16, "red"), (0.33, "red"),
                                                (0.33, "blue"), (0.49, "blue"),
                                                (0.49, "yellow"), (0.65, "yellow"),
                                                (0.65, "green"), (0.81, "green"),
                                                (0.81, "orange"), (1, "orange")])

fig.update_layout(coloraxis_colorbar=dict(
    title="Group",
    tickvals=[0,1,2,3,4,5],
    ticktext=list(grp.Group.unique()),
    lenmode="pixels", len=100,
))

# from pandas.plotting import parallel_coordinates # Make the plot vis = df[chemic_features + ['Group']] parallel_coordinates(vis.sample(100), 'Group')#, colormap=plt.get_cmap("Set2")) plt.show()

Morphology

Inspect distribution of morphological features in each group.

morph_features = ["GrayMean", "Solidity", "Irregularity", "ShapeFactor", "LengthMicron", "Eccentricity", "GrayKurt", "FractalDimR", "AreaMicron2", "GrayStd", "Grain3"]

organic_morph = organic_df[morph_features]
organic_morph["AreaMicron2"] = organic_morph["AreaMicron2"].transform('log')
organic_morph["Irregularity"] = organic_morph["Irregularity"].transform('log')
organic_morph["LengthMicron"] = organic_morph["LengthMicron"].transform('log')
organic_morph.boxplot(figsize=(18, 6))

<AxesSubplot:>

metallic_morph = metallic_df[morph_features]
metallic_morph["AreaMicron2"] = metallic_morph["AreaMicron2"].transform('log')
metallic_morph["Irregularity"] = metallic_morph["Irregularity"].transform('log')
metallic_morph["LengthMicron"] = metallic_morph["LengthMicron"].transform('log')
metallic_morph.boxplot(figsize=(18, 6))

<AxesSubplot:>

tire_morph = tire_df[morph_features]
tire_morph["AreaMicron2"] = tire_morph["AreaMicron2"].transform('log')
tire_morph["Irregularity"] = tire_morph["Irregularity"].transform('log')
tire_morph["LengthMicron"] = tire_morph["LengthMicron"].transform('log')
tire_morph.boxplot(figsize=(18, 6))

<AxesSubplot:>

mineral_morph = mineral_df[morph_features]
mineral_morph["AreaMicron2"] = mineral_morph["AreaMicron2"].transform('log')
mineral_morph["Irregularity"] = mineral_morph["Irregularity"].transform('log')
mineral_morph["LengthMicron"] = mineral_morph["LengthMicron"].transform('log')
mineral_morph.boxplot(figsize=(18, 6))

<AxesSubplot:>

grp = df[morph_features + ['Group']]
grp["AreaMicron2"] = grp["AreaMicron2"].transform('log')
grp["Irregularity"] = grp["Irregularity"].transform('log')
grp["LengthMicron"] = grp["LengthMicron"].transform('log')
grp = grp.groupby('Group', group_keys=False).apply(lambda x: x.sample(50, replace=True)).copy()
vis = grp.copy()
vis.Group = vis.Group.astype('category').cat.codes
fig = px.parallel_coordinates(vis.reset_index().drop('index', axis=1), color = "Group",
                        range_color=[0, 5],color_continuous_scale=[(0.00, "grey"),   (0.16, "grey"),
                                                (0.16, "red"), (0.33, "red"),
                                                (0.33, "blue"), (0.49, "blue"),
                                                (0.49, "yellow"), (0.65, "yellow"),
                                                (0.65, "green"), (0.81, "green"),
                                                (0.81, "orange"), (1, "orange")])

fig.update_layout(coloraxis_colorbar=dict(
    title="Group",
    tickvals=[0,1,2,3,4,5],
    ticktext=list(grp.Group.unique()),
    lenmode="pixels", len=100,
))

Distribution of sizes

import matplotlib.pyplot as plt

lengt_df = df[["Group", "LengthMicron"]]
#lengt_df["LengthMicron"] = lengt_df["LengthMicron"].transform('log')
plt.yscale("log")  
lengt_df[["Group", "LengthMicron"]].groupby('Group').boxplot(subplots=False, figsize=(15, 8))
_ =plt.xticks(rotation = 15) 

Distribution per station.

df.Sample.value_counts()

ME22_MA_P1                   1218
ME22_MA_P2                   1023
ME22_TG_P2                    982
ME22_TLY_P2                   962
BO22_Car_p1_sec. analysis     892
CA22_UV_p1_sec. analysis      871
ME22_MA_P3                    851
ME22_TG_P1                    333
ME22_TLY_P1                   288
Name: Sample, dtype: int64

_, axes = plt.subplots(3,3, figsize=(15, 8), constrained_layout=True)
axes = axes.ravel()

for station, ax in zip(df.Sample.unique(), axes):
    df[df.Sample == station].Group.value_counts().plot(kind="bar", ax= ax, title=station)
    ax.set_xticklabels(ax.get_xticklabels(), rotation = 10)

# seaborn _ ,ax = plt.subplots(figsize= (15, 8)) sns.countplot(data=df, x='Group', hue='Sample')

Proporcion de particulas por estacion

# !pip install plotly==5.9

import plotly.express as px
vis = df.groupby(['Sample', 'Group']).Score.count().reset_index(name='count')
figGeneral = px.pie(vis, values='count', names='Group', title='Poporcion de grupos por estacion', 
                    facet_col='Sample', facet_col_wrap=3)  # only plotly 5.9
figGeneral.update_layout(
    autosize=False,
    width=900,
    height=900)

import plotly
plotly.__version__

'5.9.0'

import inspect
inspect.getfullargspec(px.pie)

FullArgSpec(args=['data_frame', 'names', 'values', 'color', 'facet_row', 'facet_col', 'facet_col_wrap', 'facet_row_spacing', 'facet_col_spacing', 'color_discrete_sequence', 'color_discrete_map', 'hover_name', 'hover_data', 'custom_data', 'category_orders', 'labels', 'title', 'template', 'width', 'height', 'opacity', 'hole'], varargs=None, varkw=None, defaults=(None, None, None, None, None, None, 0, None, None, None, None, None, None, None, None, None, None, None, None, None, None, None), kwonlyargs=[], kwonlydefaults=None, annotations={'return': <class 'plotly.graph_objs._figure.Figure'>})

px.histogram(df, x="Sample", color="Group", barmode='group', title='Conteo de grupos por estacion') 

Tire - metallic

LLanta aparece con mayor proporcion en las 3 muestras de museo de antioquia seguido de las etaciones de bogota y la muestra numero 2 de tanques la Y. Esto puede indicar la presencia de alto flujo vehicular en estos muestreos. Particulas metalicas aparecen con distribucion similar en los 3 sitios teniendo un pico en la segunda muestra de tanques la Y, indicando alto flujo vehicular descendiente de la pendiente durante el tiempo de muestreo.

Minerales y organicos

Minerales aparecen con mayor proporcion en las estaciones urbanas sugiriendo resuspencion de suelos, con excepcion de tanques G2 donde se presenta un pico que puede indicar origenes de actividades de construccion. Organicos aparecen en proporciones similares en todas las estaciones mostrando un pico en tanques y2.

Salt - Ti Rich

sales y Ti Rich aparecen solo en estaciones urbanas con mayor en medida en MA donde se realizan muchas actividades de carpinteria y talleres de de autos.

Clustering

cluster particles by features with T-sne dim reduction

features = morph_features + chemic_features
#features.append("Group")

cl_df = df[features]
cl_df["AreaMicron2"] = cl_df["AreaMicron2"].transform('log')
cl_df["Irregularity"] = cl_df["Irregularity"].transform('log')
cl_df["LengthMicron"] = cl_df["LengthMicron"].transform('log')
cl_df

	GrayMean	Solidity	Irregularity	ShapeFactor	LengthMicron	Eccentricity	GrayKurt	FractalDimR	AreaMicron2	GrayStd	Grain3	Si	C	O	P	Cl	Na	Ca	Al	Fe
0	0.214	0.782	1.779181	4.220	2.793065	0.975	-1.009	1.042	3.587400	0.041	0.025	0.000	58.245	38.407	1.272	0.293	0.000	0.000	0.000	0.000
1	0.234	0.863	1.495373	2.724	2.570549	0.972	-0.985	1.013	3.294503	0.052	0.010	0.000	58.868	36.102	1.123	0.505	0.354	0.000	0.000	0.000
2	0.201	0.956	0.273837	1.141	1.545006	0.541	-0.133	1.012	2.637842	0.026	0.000	0.000	66.937	31.014	0.658	0.000	0.309	0.000	0.000	0.000
3	0.238	0.973	0.413433	1.087	1.344951	0.721	0.086	1.001	2.139125	0.042	0.000	0.000	56.770	32.504	0.956	0.000	0.000	0.000	0.000	0.000
4	0.227	0.973	0.093490	1.010	1.099612	0.267	2.696	1.001	1.936148	0.029	0.000	0.270	85.876	13.854	0.000	0.000	0.000	0.000	0.000	0.000
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
7415	0.315	0.796	0.954357	2.153	2.265714	0.767	0.673	1.015	3.507118	0.092	0.016	10.747	34.808	33.543	0.000	0.000	0.486	1.759	5.756	10.721
7416	0.300	0.917	0.509825	1.274	1.640549	0.703	-0.146	1.008	2.727918	0.081	0.023	5.470	44.663	32.910	0.221	0.000	0.298	2.212	2.763	8.667
7417	0.203	0.664	1.010873	5.696	2.006199	0.524	9.749	1.066	3.123202	0.064	0.012	9.509	36.093	31.405	0.000	0.000	0.519	0.783	4.940	11.889
7418	0.312	0.829	0.692647	1.789	1.420937	0.395	3.081	1.035	2.130966	0.084	0.032	4.181	32.476	35.681	0.000	0.000	0.508	12.860	1.610	10.304
7419	0.285	0.895	1.121678	1.648	1.565277	0.924	0.298	1.011	1.932535	0.065	0.026	6.255	51.504	26.648	0.000	0.000	0.461	0.500	3.502	8.516

7420 rows × 20 columns

#import the method
from sklearn.manifold import TSNE
#set the hyperparmateres
keep_dims = 2
#lrn_rate = 700
prp = 40
#extract the data as a cop
tsnedf = cl_df.copy()
#creae the model
tsne = TSNE(n_components = keep_dims, 
            perplexity = prp, 
            random_state = 42,
            n_iter = 5000,
            n_jobs = -1)
#apply it to the data
X_dimensions = tsne.fit_transform(tsnedf)
#check the shape
X_dimensions.shape

(7420, 2)

X_dimensions

array([[-106.354195,  -18.820627],
       [-103.93233 ,  -10.525301],
       [ -94.231544,   32.022583],
       ...,
       [ -17.325533,    5.60417 ],
       [ -18.602459,  -26.244287],
       [ -49.345295,   10.416051]], dtype=float32)

tsnedf_res = pd.DataFrame(data = X_dimensions,
                         columns = ["Dimension 1", 
                                "Dimension 2"])

tsnedf_res["Group"] = df.Group.copy()

tsnedf_res

	Dimension 1	Dimension 2	Group
0	-106.354195	-18.820627	Biogenic_Organic
1	-103.932327	-10.525301	Biogenic_Organic
2	-94.231544	32.022583	Biogenic_Organic
3	-78.440651	-7.695202	Biogenic_Organic
4	-17.732616	87.539787	Biogenic_Organic
...	...	...	...
7415	-12.658135	-7.580052	Tire wear
7416	-37.999176	-4.604805	Tire wear
7417	-17.325533	5.604170	Tire wear
7418	-18.602459	-26.244287	Tire wear
7419	-49.345295	10.416051	Tire wear

7420 rows × 3 columns

import seaborn as sns

#plot the result
g = sns.jointplot(data = tsnedf_res,
                 x = "Dimension 1",
                 y = "Dimension 2",
                 hue = "Group")

Train models based on features

cl_df["Group"] = df.Group.copy()
cl_df['Score'] = df.Score.copy()

Binary models

import statsmodels.api as sm
import numpy as np

data = cl_df.iloc[:, :-2]
data = data.join(pd.get_dummies(cl_df.Group))

Biogenic_Organic

X = data.iloc[:, :-6]
y = data['Metallic']

X = sm.add_constant(X, prepend = False)
lg= sm.Logit(y, X)
result = lg.fit()
print(result.summary())

Optimization terminated successfully.
         Current function value: inf
         Iterations 13
                           Logit Regression Results                           
==============================================================================
Dep. Variable:               Metallic   No. Observations:                 7420
Model:                          Logit   Df Residuals:                     7399
Method:                           MLE   Df Model:                           20
Date:                Wed, 02 Nov 2022   Pseudo R-squ.:                     inf
Time:                        20:57:49   Log-Likelihood:                   -inf
converged:                       True   LL-Null:                        0.0000
Covariance Type:            nonrobust   LLR p-value:                     1.000
================================================================================
                   coef    std err          z      P>|z|      [0.025      0.975]
--------------------------------------------------------------------------------
GrayMean        15.3646      4.494      3.419      0.001       6.557      24.172
Solidity        -6.3940      6.153     -1.039      0.299     -18.454       5.666
Irregularity    -1.6327      2.147     -0.760      0.447      -5.842       2.576
ShapeFactor      0.1312      0.199      0.660      0.510      -0.259       0.521
LengthMicron    -1.5961      3.745     -0.426      0.670      -8.936       5.744
Eccentricity     1.8351      1.797      1.021      0.307      -1.687       5.357
GrayKurt         0.0134      0.115      0.116      0.908      -0.213       0.239
FractalDimR     14.9378      9.844      1.517      0.129      -4.357      34.233
AreaMicron2      0.4625      1.906      0.243      0.808      -3.274       4.199
GrayStd         25.5181      6.130      4.163      0.000      13.503      37.533
Grain3           1.5785      6.751      0.234      0.815     -11.653      14.810
Si              -0.2225      0.041     -5.415      0.000      -0.303      -0.142
C               -0.1194      0.020     -5.842      0.000      -0.159      -0.079
O               -0.4178      0.053     -7.878      0.000      -0.522      -0.314
P               -0.3466      0.132     -2.621      0.009      -0.606      -0.087
Cl              -0.1973      0.265     -0.746      0.456      -0.716       0.321
Na              -0.2062      0.270     -0.765      0.444      -0.735       0.322
Ca              -0.1192      0.033     -3.578      0.000      -0.185      -0.054
Al              -0.1881      0.056     -3.349      0.001      -0.298      -0.078
Fe               0.2789      0.031      8.984      0.000       0.218       0.340
const           -5.5269     14.164     -0.390      0.696     -33.288      22.234
================================================================================

Possibly complete quasi-separation: A fraction 0.75 of observations can be
perfectly predicted. This might indicate that there is complete
quasi-separation. In this case some parameters will not be identified.

/home/lecun/.local/lib/python3.8/site-packages/statsmodels/base/model.py:547: HessianInversionWarning:

Inverting hessian failed, no bse or cov_params available

/home/lecun/.local/lib/python3.8/site-packages/statsmodels/base/model.py:547: HessianInversionWarning:

Inverting hessian failed, no bse or cov_params available

from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(cl_df.iloc[:, :-2], cl_df.Group, test_size=0.30, random_state=42)
print(len(X_train), len(X_test))

5194 2226

from sklearn.linear_model import LogisticRegression

sklg = LogisticRegression(multi_class='multinomial', solver='lbfgs', penalty='none')
sklg.fit(X_train, y_train)
pred = sklg.predict(X_test)

from sklearn.metrics import ConfusionMatrixDisplay, classification_report, confusion_matrix
print(classification_report(y_test, pred))

                  precision    recall  f1-score   support

Biogenic_Organic       0.95      0.97      0.96       719
        Metallic       0.97      0.91      0.94       136
         Mineral       0.95      0.97      0.96      1109
            Salt       0.50      0.25      0.33         4
   Ti-rich Paint       0.00      0.00      0.00        19
       Tire wear       0.85      0.78      0.82       239

        accuracy                           0.94      2226
       macro avg       0.70      0.65      0.67      2226
    weighted avg       0.93      0.94      0.93      2226

confusion_matrix(y_test, pred)

array([[ 697,    0,    7,    0,    2,   13],
       [   2,  124,    9,    0,    1,    0],
       [   7,    3, 1080,    1,    0,   18],
       [   1,    0,    2,    1,    0,    0],
       [  18,    0,    0,    0,    0,    1],
       [  12,    1,   39,    0,    0,  187]])

ConfusionMatrixDisplay.from_predictions(y_test, pred, cmap='Blues')

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7f4719f64550>

Test Deep Model

#Run this command in the appropied environment. 
!torchrun classification/train.py  --test-only --model=resnet18 --data-path=/media/lecun/HD/Expor2/ParticlesDB/folders --resume=./classification/models/model_6.pth

Loading data
Loading training data
Took 0.011496543884277344
Loading validation data
Creating data loaders
/media/lecun/HD/Expor2/Particle-classifier/classification/env/lib/python3.8/site-packages/torch/utils/data/dataloader.py:563: UserWarning: This DataLoader will create 16 worker processes in total. Our suggested max number of worker in current system is 8, which is smaller than what this DataLoader is going to create. Please be aware that excessive worker creation might get DataLoader running slow or even freeze, lower the worker number to avoid potential slowness/freeze if necessary.
  warnings.warn(_create_warning_msg(
Creating model
Test:   [ 0/59]  eta: 0:01:47  loss: 0.1661 (0.1661)  acc1: 96.8750 (96.8750)  acc5: 100.0000 (100.0000)  time: 1.8303  data: 1.0699  max mem: 257
Test:  Total time: 0:00:03
Test:  Acc@1 84.797 Acc@5 99.679
[[598   2 111   2]
 [  2  94  17   1]
 [  9  17 753  51]
 [  7   0  65 139]]
                  precision    recall  f1-score   support

Biogenic_Organic       0.97      0.84      0.90       713
        Metallic       0.83      0.82      0.83       114
         Mineral       0.80      0.91      0.85       830
       Tire wear       0.72      0.66      0.69       211

        accuracy                           0.85      1868
       macro avg       0.83      0.81      0.82      1868
    weighted avg       0.86      0.85      0.85      1868

cm = [[598,2,111,2],
 [2,94,17,1],
 [9,17,753,51],
 [7,0,65,139]]

disp = ConfusionMatrixDisplay(confusion_matrix=np.array(cm))
disp.plot(cmap='Blues')

<sklearn.metrics._plot.confusion_matrix.ConfusionMatrixDisplay at 0x7f46b42d3c70>